HAL Id: inria-00496873
https://hal.inria.fr/inria-00496873
Submitted on 1 Jul 2010
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
TERNARY QUARTIC APPROACH FOR POSITIVE 4TH ORDER DIFFUSION TENSORS REVISITED
Aurobrata Ghosh, Rachid Deriche, Maher Moakher
To cite this version:
Aurobrata Ghosh, Rachid Deriche, Maher Moakher. TERNARY QUARTIC APPROACH FOR POS-
ITIVE 4TH ORDER DIFFUSION TENSORS REVISITED. 2009 IEEE International Symposium on
Biomedical Imaging: From Nano to Macro, Jun 2009, Boston, United States. 4 p. �inria-00496873�
TERNARY QUARTIC APPROACH FOR POSITIVE 4TH ORDER DIFFUSION TENSORS REVISITED
Aurobrata Ghosh, Rachid Deriche
INRIA Sophia Antipolis-M´editerran´ee Project Team Odyss´ee
Sophia Antipolis, France
Maher Moakher
Ecole Nationale d’Ing´enieurs de Tunis (ENIT) ´ LAMSIN
Tunis, Tunisia
ABSTRACT
In Diffusion Magnetic Resonance Imaging (D-MRI), the 2nd order diffusion tensor has given rise to a widely used tool – Diffusion Tensor Imaging (DTI). However, it is known that DTI is limited to a single prominent diffusion direction and is inaccurate in regions of complex fiber structures such as crossings. Various other approaches have been introduced to recover such complex tissue micro-geometries, one of which is Higher Order Cartesian Tensors. Estimating a positive diffusion function has also been emphasised mathematically, since diffusion is a physical quantity. Recently there have been efforts to estimate 4th order diffusion tensors from Diffusion Weighted Images (DWIs), which are capable of de- scribing crossing configurations with the added property of a positive diffusion function. We take up one such, the Ternary Quartic approach, and reformulate the estimation equation to facilitate the estimation of the non-negative 4th order diffu- sion tensor. With our modified approach we test on synthetic, phantom and real data and confirm previous results.
Index Terms — Diffusion-MRI, Higher Order Tensors, Ternary Quartics, Diffusion Propagator
1. INTRODUCTION
Diffusion Magnetic Resonance Imaging (D-MRI) provides a sophisticated tool to study the connectivity of the brain’s white matter in vivo. This non-invasive approach makes it possible to determine the micro-structure of the tissue by measuring and quantizing the diffusion of water molecules in a restricted environment. This allows to infer the underly- ing geometry. More specifically partial directional diffusion information contained in multiple Diffusion Weighted Im- ages (DWIs) are integrated using a reconstruction model to generate an image where at every voxel a diffusion function indicates prominent diffusion directions which reflect the geometry of the tissue and point out important fiber bundles.
The earliest proposed diffusion function employing a 2nd order Cartesian tensor has given rise to Diffusion Tensor Imaging [1], the most popularly utilised approach today. The
diffusion function was defined as D(g) = g
TDg, where g is the diffusion weighting magnetic gradient vector and D is the 2nd order tensor to be estimated from a set of DWIs. It is well known, however, that the 3D, 2nd order tensor is incapable of describing more than one prominent fiber direction, and thus is inaccurate when complex fiber configurations are present such as crossings.
In [2], Higher Order Tensors (HOT) were introduced to allow for a more complex diffusion function capable of de- scribing crossing fiber configurations. The diffusion function was remodelled to employ a 3D, Cartesian HOT and written as
D(g) =
3j1=1
3j2=1
· · ·
3jk=1
D
j1j2...jkg
j1g
j2. . . g
jk, where g
jiare again the coordinates of the magnetic gradient vector and D
j1j2...jkare now the independent coefficients of the k-th order diffusion tensor.
Since diffusion is a physical quantity it is meaningless if the diffusion function is negative in any direction. The authors of [3] worked with a 4th order diffusion tensor with this added constraint of non-negative diffusion. For the 4th order they rearranged the tensor indices to write the diffusion function as
D(g) =
i+j+k=4